I know in France this system have been used by Decathlon (which is a really appreciated chain store of sports equipment) for multiple years.
I’m not sure if they developed it themselves (I know they do have some tech team but I don’t know if they have the capacity to develop hardware) or if it’s an off the shelf solution. It’s also possible that the needed hardware parts were already available and that they developed the software and the chassis.
And yes this tech is like magic, you just throw all your stuff in the bin and boom, you pay.
I also suspect that it only works for baskets that are not too dense (so clothes and sport equipment are ok) because otherwise we would have this in every grocery store since the gain in time is tremendous even compared to self-scan cashiers. It’s basically longer to scan your loyalty card and to contactless pay than to scan your 25 articles.
Or maybe it’s just the cost of RFID tags that prevents this to happen because at the end of the day, queues at cashier are rarely a money bottleneck.
Two things prevent this tech from being deployed by more stores.
1. Cost of RFID tags, they’re very cheap, but still an order of magnitude more expensive than barcodes and have a meaningful impact on item costs in some markets (notably groceries).
2. Supply chain problems. This is actually a much bigger issue than the cost. RFID based checkouts require a store of have every single one of their suppliers integrated the tags into their products. Most shops simply don’t have enough control of their supply chains to make that happen in a reasonable timeframe. Until every product on the shelves has tags, you can’t deploy new checkouts, but you still need to pay for RFID tags you can’t use during the migration.
Decathlon have RFID based checkouts because they mostly sell their own products they’ve designed and had manufactured. For the vast majority of their products the RFID tag is stitched directly into the product, so it’s not a label that can be easily peeled off, or fall off. For the small number of 3rd party products they sell, they’ve got enough volume to get other suppliers to integrate RFID tags, or they can use removable labels because the risk of lost labels is acceptable if it only applies to 1% of your stock.
So don’t expect to see this type of tech deployed elsewhere anytime soon. Decathlon’s extremely strong vertical integration makes it substantially cheaper and easier for them to deploy RFID, compared to most other ships.
> RFID based checkouts require a store of have every single one of their suppliers integrated the tags into their product.
I'm sure the same was said about the introduction of barcodes in the 1970's. How can every single manufacturer be expected to adopt a uniformed standard, where there are not clashes between standards and adoption? Well, they managed it by collaboration: https://en.wikipedia.org/wiki/GS1
The same was also said about chip-and-pin payments, and further with online payments. This was largely implemented through regulation by the EU.
Barcodes are much cheaper to implement, and provided a much greater benefit compared to the alternative (no automatic way of recognising goods), compared to RFID. RFID improves the customer UX experience at self checkouts, and also the opportunity to improve stock management (stock counts could be reduced to just waving an antenna over a shelf). But that improve is pretty marginal compared to barcodes. Also RFID tags aren’t a simple to introduce as barcodes, you can just slap an RFID tag on existing packing and just expect it to work, there are issues when dealing products that contain metal etc. Not insurmountable, but one more fly in the ointment.
Chip-and-pin is very poor comparison, there’s realistically only 3 major card networks in Europe. If they mandate Chip-and-PIN, then chip-and-PIN happens. They only need to update card readers, of which there is only a small number of manufacturers, and they all already are used to following strict standards, and cards can be replaced as they expire, and again only a small number of manufacturers.
Simply put RFID requires a huge number of manufacturers to correctly implement RFID tags on all their products, just so a single store can implement better checkouts. Why would manufacturers voluntarily pay for that cost when there’s no clear benefit for them? For stores, why would want to cover the cost of their suppliers adding RFID infrastructure to their supply chain, when the benefits are fairly limited compared to barcodes?
That’s completely to Chip-and-PIN, where only the stores need to update their readers, and they can easily change to which ever manufacturer is willing to make them.
Companies like Decathlon and Uniqlo have a huge advantage, because they get huge supply and stock control benefits from RFID due to their vertical integration. Better self checkout are just the cherry on top, but won’t be the primary reason for adopting RFID tech.
It is more than just improving the customer UX, it creates massive savings for the retailers. The further reduction of manned checkouts, and the faster customer throughput (some data that it is 25x faster) meaning reduced queues, which increases volume of sales. Every bit of reduced friction for the customer, increases profit potential.
Further, loss prevention is reduced as accidental scanning errors, and ease of intentional theft is reduced. The issue you raise of containing metal is a historic problem, but has been solved for many years through RAIN RFID tags, where a metal surface itself actually becomes part of the antenna, making it even stronger. Are you aware of any further current issues?
I agree, the cost is higher than a barcode, but when the barcode was introduced the same argument was valid. What is the benefit to the manufacturer to get involved in stock management, when the shop was previously responsible for "pricing-up" the item themselves. Wrigley's gum was the first to do it, but how did it benefit them?
With more adoption, the price will decrease, and this will become standard. Or, what is the alternative? That we use barcodes for the next 100 years?
Regarding chip-and-pin, I was responsible for writing some of the middleware in ~2004 that interfaced between chip-and-pin machines and terminals (& other devices), for what was then the largest provider in this space. It's wild that you state it was as simple as you claim, it was a massive effort across the industry, requiring lots of revisions to standards, collaboration, conferences etc. You feeling it was just a case of swapping the machine over, my former colleagues and I should take as a complement that we did it so well to give that impression.
Something that occurs to me might be a problem in grocery stores is vegetables sold by weight. I can purchase potatoes, carrots, tomatoes, and other vegetables by weight and the scale will print a barcode to be scanned. Can one just print off an RFID tag like that? Seems unlikely.
My prediction is that the ability to choose-your-own will go away over the next 10 years, and it will only be the option to buy pre-packaged. For the retail outlet, weigh-your-own is slow to process and a lot of wastage.
Which is a shame, because it is better for the consumer and nearly always cheaper for us to choose-our-own.
> The issue you raise of containing metal is a historic problem, but has been solved for many years through RAIN RFID tags
Yes I know it’s solved. But it’s not as simple as barcodes. Each product needs to be evaluated to decide if you’re gonna put a tag on it that works well on metal surfaces, because those tags tend to be marginally more expensive than the alternatives.
> I agree, the cost is higher than a barcode, but when the barcode was introduced the same argument was valid. What is the benefit to the manufacturer to get involved in stock management, when the shop was previously responsible for "pricing-up" the item themselves.
Simple, there’s an obvious cost reduction for shops, no more need for labels or hours of pricing up, makes it easier for a store to eat the costs for their manufacturers. For manufacturers, the improved stock management at stores would allow this store to carry a large variety of brands, which makes it easier for manufacturers to get their products onto shelves.
> With more adoption, the price will decrease, and this will become standard. Or, what is the alternative? That we use barcodes for the next 100 years?
The price of RFID tags isn’t expected to get any lower. General consensus is that there’s already enough demand for RFID tags that they’ve more or less hit their lower possible price, and any further reduction are not going to be meaningful.
What’s wrong with using barcodes? It’s a well proven technology, you could update them to 2D barcodes if you want more encoded data. There’s no reason to get rid of technology that works perfectly well just for the sake of using a “newer” technology. Most retailers are looking at Amazon fresh stores and wondering if that can be made to scale to grocery stores.
> Further, loss prevention is reduced as accidental scanning errors, and ease of intentional theft is reduced.
The types of thefts you’re referring to aren’t meaningful compared to theft carried out by organised crime, and by the staff themselves. RFID tags aren’t going to stop people walking into stores and simply picking up and entire rack of clothes, or sweeping all stakes in the meat section, and walking out. That kind of theft is a real problem, and store staff deliberately don’t stop those thieves because the risk of violence is far too high.
> Regarding chip-and-pin, I was responsible for writing some of the middleware in ~2004 that interfaced between chip-and-pin machines and terminals (& other devices), for what was then the largest provider in this space.
I’m very aware of the complexities involved in any kind of change to the EMV standards, or simple changes in payment network rules. I’ve been involved in both the negotiations and technical implementation of those changes. The difference is the number of parties and teams involved. A single grocer will have hundreds of suppliers each, and likely thousands of manufacturers. The number of parties involved is substantially greater than the number of issuers and acquirers in the world. So yes, I do think stuff like Chip-and-PIN was simpler to implement than RFID. It’s all relative.
Walmart tried to use EPC tags (UHF long-range tags) a long while ago on apparel. It died out, but EPC is 100% back.
Target and Walmart both use EPC tags now as part of the inventory journey and it wouldn't surprise me to see more and more integration down the line. Look for the GS1 EPC logo on stickers on things at Walmart...
You also have to be sure that the RFID tags can't be easily removed from the items. Otherwise, people would just remove the RFID tags, stuff the items in the bin, get told that they don't owe anything, and walk out scot free.
But how do you do that with blueberries. Does each blueberry have to have its own unique non-removable RFID tag?
There's all sorts of items at a grocery store where it's hard to guarantee the tags can't get removed from the items. That's part of why Amazon doesn't use RFID tags in their "just walk out" stores.
Also anti-theft devices. I just saw a tiktok where Zara had a similar RFID self-checkout system, but the customer had to remove the anti-theft dongles themselves. It makes for a hugely different customer experience.
How would a grocery store look that's designed around that? Sell only essential goods where you can control the supply chain, and sell it only in big portions so that the cost of the RFID tag doesn't matter.
Shops like Aldi are good place to look. They still use barcodes, but they’ve got their manufacturers to print the barcodes multiple times on packaging, and print them huge[1]. This is done to speed up scanning at their checkouts, so cashiers don’t need to look for the barcode, they just grab the item in any orientation and yank it across the scanner. That in turn means they get higher throughput per cashier, which means fewer cashiers and tills.
They also tell the cashiers that their family will be killed if they don't process each item at at least Mach 1. (Or that's what it feels like at the checkout anyway...)
Aldi is German, and “grocery packing” is the unsung national sport, everybody is at it, all the time.
You have not know humiliation if you have not experienced a line of slightly elderly shoppers mildly frowning and tutting as you hurl your items into a bag, whilst the cashier looks at you with overt disapproval. The only accepted bagging speed is “Superhuman” and the best day of my life was when the cashier at my regular Edeka nodded approvingly at the fact that I was practically ripping the groceries out of their hand as soon as the cash register beeped.
Supermarket staff here in Italy, where I now reside, think I am insane, and I work hard to contain my impatience with those lesser-trained people in front of me at the cash register.
Cashiers are measured on their scan speed, and get bonus based on performance. So yeah, they’re highly incentivised to scan as fast as they physically can.
It goes too far though. They're throttled by how fast the customer can bag things, so going way faster and overloading the packing area is counterproductive. (+ the experience makes me that much less likely to choose that shop again in the future)
It's the cost of RFID tags that makes this not viable for grocery. Grocery has much lower value per item so the cost for each tag is prohibitive. Last time I ran that math it would wipe out the entire margin of the business. Grocery retail operates by huge volume over very small per-item margins.
I might be able to share some insights on this one : I worked on the uses-cases, tech design and first deployments of RFID at decathlon, back in… 2008-2009 I guess.
We faced major issues back then, but the most important one was the cost of tagging every single product ; both cost of the tag and cost of attaching it (comparatively, antennas and IT where marginal costs at scale). Back in the end of the 2000s, tags were more expensive than today. Decathlon had one major advantage though : it manufactures most of the product. That means we could add the tag in the label at the factory. But still, to make it economically viable, we had to pile-up multiple uses cases. Every single one was important in the balance to justify buying so many tags and tagging so many products. So we went very creative, designing new ways to reduce costs in the whole manufacturing/logistics/storage/retail/after-life chain. Because no single advantage was enough by itself. We also had to imagine these new enablers and uses-cases with the state of this new technology (especially at the time) : airwaves are difficult to predict, and not 100% perfect in terms of detection, collision, …
We invented new ways to do things that were mostly already done, but cheaper (in man-hours for repetitive tasks mostly), better (quality and fiability of results), and/or more frequently. Thing logistics optimization (wrong door / parcel detected earlier), whole lifecycle and whereabouts tracking (for defective product callbacks for example), daily inventory (we had the idea to use cleaning carts with antenas to have a daily store inventory), NOSBOS (« not on shelves but on stock », detecting prodcts that where in inventory but not available at this specific moment in the department/shelf), warranty tracking, …
At the end of the story, self-checkout is mostly a by-product, not a target for the deployment. It’s something that was made possible but not a first target.
Also in use at (at least some of) Paris’ libraries. I was quite impressed with the frictionless book borrowing experience. Scan your card and put your mess of a pile of kids books on the platform. Done. Happy kid.
To boot, I was worried I was taking too many books. Expected a limit of 5 or maybe 10. But when I asked —- 40!
I first saw this system at a Texas Instruments demo in the UK in 2007-ish. They also demoed walking through a RFID reading gate with a trolley full of tagged goods.
I thought it was very cool at the time but I assumed it must have been too impractical for some reason until Decathlon rolled it out in 2019.
Went to Europe a few months ago and experienced this checkout method in Uniqlo and Zara. It was mind blowing at first and I was trying to figure out wtf was going on. Was it AI?, maybe some complex algorithm based on item weight and other variables?
Turns out they were “simply” RFID chips. It’s amazing how magical technology can be. I hope the tech spreads because I hate the checkout experience.
So there are disposable RFID chips on every price tag? I would think that is wasteful and expensive? Is there a way to recapture them like a deposit return
The tags are usually stitched directly into the clothes. Realistically they’re never going to be recaptured and returned. They’ll be recycled/disposed of, when the garment itself reaches EOL.
I think they're even on the price tags of items you buy online. I've noticed that my Uniqlo online orders also have some kind of circuitry inside them.
I've seen bookstores implement disposable RFID for loss prevention more than a decade ago where the books sell for around as much as these garments. I suppose the marginal cost of RFIDs that support this system is small if you already intend to use disposable RFIDs for loss prevention, with a large limiting factor being the acceptance of self-checkout.
some clothes come with the tag sewed in from the factory (not talking about uniqlo). which is worrisome because the combination of rfid tags you wear could be used to track you
According to a relevant study ([1]), "Environmental Burden Case Study of RFID Technology in Logistics Centre", the mean weight of the metal with foil is 0.161 g per tag (see Table 2).
Let's suppose that every person on Earth (~8B) buys 1000 RFID-encoded goods per year: 0.161 * 8B * 1000 = 1.3e12 g = 1.3e9 kg = 1.3 M tonnes. In 2022, some 20M tonnes of copper has been mined ([2]), which gives us ~5% of the total copper produced. If aluminum is used, then ~2%.
Takeaways:
- yes, it could be a lot of metal, if the RFID technology gets widespread (narrator voice: it will)
- it will visibly increase demand for aluminum / copper, but it won't be the end of the world.
People really tried to make RFIDs work, but it's a poor fit for anything with metal in it (cans of beans, electronics, etc.). Sun and IBM tried to push them in 2000-s (remember that ad: https://www.youtube.com/watch?v=k3Fdox5_rg4 ), and it never worked in real life.
Uniqlo checkouts work because they are selling mostly, well, clothes that you can just put inside the scanner.
I think key here is that it's 100% recyclable or even reusable. The used tags could even be picked up at the exit/checkout if I understand correctly the process correctly.
The retailer can add RFID labels. They do this in decathlon (sporting goods store in Europe) where they have used same checkout process for a few years
No they dont. They sell many 3rd party brands in store in nearly every category. I believe they were (mostly) own-brand up until a few years ago, but thats certainly not the case now.
They are chips. Press release suggests they're using Impinj's ISO-18000-63 chips. Those chips are small (datasheet says 362×247×120 microns) and directly bonded to the printed antenna inlay.
Like others in the thread, I also experienced this recently in NYC. It’s amazing how incredibly quick the system is too. You place the items in the bin at the same time, and by the time you look up at the screen the price is already waiting for you to confirm and proceed to payment.
One of the first times I’ve felt delight shopping in person in years.
My local library system has had this for years. I’ve always wondered why it’s not more ubiquitous. You put your stack of books on the counter, and the titles appear on a confirmation screen almost instantaneously. Scan your library card and away you go.
A library book is a relatively expensive object (not necessarily the physical book, but the whole cost of having the book at the library). Spending a couple dollars on a durable tag that is permanently attached to the book makes sense.
Retail stores had to wait until people developed much cheaper tags, because each tag only gets used once. These cost a few cents per tag, and are different from the 13.56MHz tags that the library probably uses.
Hm, you're right, even the 13.56 tags have come way down in price. Anyway, I think my point still stands: RFID as a technology used to be more expensive, and used to be at a price point were tracking an asset like a book or a shipping crate through multiple uses made sense but tracking a bottle of milk through a single purchase didn't.
Someone called me for an interview with Amazon. They were working on an image recognition system for physical stores. Once you got an item from the shelve and put it on the basket they've open a virtual basket and add the item there, too. Once you get to checkout, all you have to do is pay.
I had a similar idea when I developed checkout for Metro Cash and Carry. Put some cameras near the wait line at the cash registers and use image recognition instead of having a clerck scan all the goods. They weren't interested, though.
This is called amazon go. I believe they had intentions of selling this tech to other stores but it never really took off (maybe uniqulo is using it?). They closed several of Amazon's stores which used this in 2023 and there are only around 40.
Just put the “self checkout” on the cart/trolley and take them directly to your car. “SmartCarts” have to be a thing at some point. …and they will put video ads on them.
Amazon Fresh grocery stores do something similar, but with the tech (cameras, maybe augmented by weight/pressure sensors?) in the shopping carts rather than covering the ceiling.
From what I’ve gathered talking to people who work there about 80-90% of the tech is a gig worker in Africa or Asia diligently watching a bunch of cameras and typing stuff in that gets rebranded as AI.
> They closed several of Amazon's stores which used this in 2023
I think this is partly due to WFH and loss of foot traffic (these were usually located in business districts, where the lunch crowd would come in and grab a prepared meal and walk out). I don't think it was due to technology not working or uptake being low.
Downtown Chicago had a few that were fairly popular. COVID killed the office crowd.
Basic Amazon stores never used this. Only the amazon go convenience stores ever did. As far as I can tell, the former are all closed now while many of the latter are still open. At least here in Seattle.
With Amazon Go you can request a refund for any item, whether you actually bought it or not, and they just… give it to you. Don’t even have to return the item to the store. I assume there are some limitations and it’s based on good faith, but I don’t know because I’ve never had any reason to request one.
I have had them miss items and fail to charge me for them, though. Even if you want to be honest about it though, there’s really nothing that can be done, since they don’t charge you and provide you with a receipt until hours after you’ve left the store.
Why would you put yourself through it though? It's stressful. I spend enough of my time on the phone to Amazon through AWS and because my stuff is stolen.
Stressful? In what way is it possibly stressful? Because you have to keep an eye out for a receipt where you maybe were overcharged for something (but probably weren’t) so that you can receive a guaranteed refund to correct the situation upon request?
There are a lot of ways I’d describe the Amazon Go shopping experience. Creepy? Yes. Unsettling? Quite so. Oddly evoking the feeling of being a rat in a maze? I wouldn’t argue. But stressful? If there is one thing it is not, it’s that.
Or overindexing on biased data points might cause biased conclusions, without considering standard variation.
Lots of people use it without any issues -- if something isn't truly a dealbreaker, the "I won't bother" crowd tends to become a minority. The Law of Large Numbers will work as expected.
The evolution of this is fully automated checkout like in Amazon Fresh stores.
It's fantastic: I can go to the store, shove some groceries in my pockets or backpack, and walk out in 45 seconds. Amazon pings you a notification telling you how long you were there after you leave, and I'm convinced they do it to have people try to improve their shortest times.
It seems to be a fair amount of pfaff getting in to confirm who you are, there didn’t seem to be that much selection to find much worth buying.
But automated tills have long queues, and I’ve noticed I sometimes bodge the payment system and leave without paying. But being in the habit of checking my bank account whilst shopping I notice payment hasn’t gone through and put my shopping through.
Which is strange since someone ran after me when I absent Monday walked out of a store with an empty shopping basket.
> It seems to be a fair amount of pfaff getting in to confirm who you are
This seems overstated. You scan a barcode on your Amazon app. That's it.
> there didn’t seem to be that much selection to find much worth buying.
It feels like a smaller, more focused grocery store (something between an Aldi and a Safeway), but all the basics are there and then some. I don't live close enough to one to use it regularly, but I would if I did.
This has not happened to me ever. I've picked up and put back items a ton. In fact the selling point of the tech is it is meant to handle this specific scenario. There are also weight sensors, so it's not just vision.
This doesn't work with groceries because the tags can't be read through liquids, right? Any solutions? It would be delightful to just push my cart through a scanner and then pay.
Yes. The last time this idea came around, RFID tags that could be read in bulk without interference cost about a dollar. Now they're US$0.04, says the article.
Each RFID tag needs to be queried separately. If two of them send at the same time, neither gets received. So there's a recent and clever polling scheme which sends polls to which only certain tags reply, and tags also reply in multiple time slots chosen semi-randomly. It's a cross between two old ideas - slotted ALOHA, and the scheme used to make plug-and-play work for ISA cards.
This is what allows just dumping all the stuff into a box and scanning all the tags at once.
Uniqlo sells mostly Uniqlo stuff, so they have control over tagging. Also, clothing tends to be RF-transparent. If you're selling canned goods, this might not work.
I'd say another challenge with grociers is that each item is of such differnt size, shape, material. You have tiny flexible plastic bags, bananas, basil plants, round glass beer bottles, self serve orange juice machines operated by third party companies, the meat and deli counter and so forth.
It‘s true that one forgets the well-honed hate of self-checkout lines in Uniqlo and Decathlon these days.
I find it amusing how a WSJ reporter could be so amazed by the tech. This has been the standard for years in cheap stores like Decathlon, it really doesn’t feel anything like magical.
Although I like to shop at Uniqlo, I have to say, this is not unique to them. In fact, I saw it first at Decathlon here in UK, before Uniqlo introduced theirs.
Only issue is that I am not sure they are recycleable anymore. But it is a small strip of cardboard and metal - so probably doesn't move the needle anyway.
Could be a slippery slope. It is just a plastic cup, a straw, a thin piece of metal. Overtime those can build up especially if the volume rises. Should RFID purchasing in this manner really take off, it could be like plastic all over again. Which would create another environmental concern: Mining the metal for the tags.
TL;DR: The eliminated labour they are talking about is the physical act scanning the bar code of each piece individually - instead, they are using RFID, so you can just stuff all the pieces you're buying into a bin and the machine will detect everything automatically.
I used that kind of self-checkout before and honestly find it less revolutionary than they make out to be here. It's still a hassle and there are lots of tasks which are still "transferred" to the shopper, e.g. folding up all the clothes correctly and making sure that everything is in order.
"Fast Retailing began testing RFID tags in 2013, introduced the self-checkout machines in 2014 and switched to the current model in 2019. They came to America on a trial basis in 2021, followed by an official rollout in 2022. This was the year they arrived in every Uniqlo store in the U.S."
quite - the hyperbole of the subtitle (or it's presence on HN) doesn't match the banality reality of 'checkouts used by millions of people around the world still in use'
